启发式推理先验助力数据高效的指代目标检测
Heuristic-inspired Reasoning Priors Facilitate Data-Efficient Referring Object Detection
把由指代短语推导出的空间/语义启发式先验注入 DETR 式 grounding 流程的 proposal ranking、prediction fusion 与 Hungarian matching 三个环节，是一种模型无关、可直接复用的低数据量增强手段。
原文依据 injects explicit, heuristic-inspired spatial and semantic reasoning priors
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论大多数指代目标检测（ROD）模型，尤其是现代视觉定位检测器，都是为数据充足的条件设计的，然而许多实际部署场景，如机器人、增强现实及其他专门领域，会面临严重的标签稀缺问题。在这种情形下，端到端的视觉定位检测器需要从零学习空间与语义结构，浪费了宝贵的样本。我们提出一个简单的问题：当数据稀缺时，显式的推理先验能否帮助模型更高效地学习？为探究这一问题，我们首先引入数据高效指代目标检测（De-ROD）任务，它是一个用于衡量低数据与少样本设定下 ROD 性能的基准协议。随后我们提出 HeROD（启发式启发的 ROD），一个轻量、模型无关的框架，它将显式的、受启发式驱动的空间与语义推理先验（即基于指代短语推导出的可解释信号）注入现代 DETR 式流程的三个阶段：候选排序、预测融合与匈牙利匹配。通过使训练与推理都偏向合理的候选，这些先验有望提升标签效率与收敛性能。在 RefCOCO、RefCOCO+ 与 RefCOCOg 上，HeROD 在标签稀缺情形下持续优于强视觉定位基线。更广泛地说，我们的结果表明，整合简单、可解释的推理先验为更好的数据高效视觉-语言理解提供了一条实用且可扩展的路径。代码见：https://github.com/xuzhang1199/HeROD。
英文原文摘要 1551 字符
Most referring object detection (ROD) models, especially the modern grounding detectors, are designed for data-rich conditions, yet many practical deployments, such as robotics, augmented reality, and other specialized domains, would face severe label scarcity. In such regimes, end-to-end grounding detectors need to learn spatial and semantic structure from scratch, wasting precious samples. We ask a simple question: Can explicit reasoning priors help models learn more efficiently when data is scarce? To explore this, we first introduce a Data-efficient Referring Object Detection (De-ROD) task, which is a benchmark protocol for measuring ROD performance in low-data and few-shot settings. We then propose the HeROD (Heuristic-inspired ROD), a lightweight, model-agnostic framework that injects explicit, heuristic-inspired spatial and semantic reasoning priors, which are interpretable signals derived based on the referring phrase, into 3 stages of a modern DETR-style pipeline: proposal ranking, prediction fusion, and Hungarian matching. By biasing both training and inference toward plausible candidates, these priors promise to improve label efficiency and convergence performance. On RefCOCO, RefCOCO+, and RefCOCOg, HeROD consistently outperforms strong grounding baselines in scarce-label regimes. More broadly, our results suggest that integrating simple, interpretable reasoning priors provides a practical and extensible path toward better data-efficient vision-language understanding. Code at: https://github.com/xuzhang1199/HeROD.